By Elizabeth Culliford

(Reuters) – Meta Platforms Inc, Facebook’s owner, announced Tuesday that it is offering access to a vast language model for AI research.

Meta said its model was the first 175-billion-parameter language model to be made available to the broader AI research community.

“Large Language Models” refer to natural language processing systems that have been trained with large amounts of text. These models can be used for reading comprehension and generating new texts.

Meta wrote in a blog posting that the Open Pretrained Transformer (OPT-175B ) model release would help researchers better understand large language models.

Meta claimed that access restrictions to these models have been hindering efforts to increase their robustness as well as to mitigate known problems such bias or toxicity.

The key area of artificial intelligence technology research and development is for many major online platforms. It can be used to perpetuate human societal biases about issues such as race and gender. Many researchers are worried about the dangers that large language models can spread.

Meta claimed it “hoped increase the diversity in voices that define ethical considerations for such technologies.”

According to the tech giant, it released the model with a noncommercial license in order to concentrate on research uses.

Meta indicated that the model would be made available for academic research and those associated with governments, civil societies, academic and industrial organizations as well as to industry researchers. This release will provide access to the model and instructions on how to train them and how they can be used.